6.433 Recursive Estimation

6.435 System Identification

Spring 2002

Sanjoy Mitter, Lecturer

Prerequisites: 6.241 or equivalent; 6.432 or equivalent.

The two courses will be taught as a single course and will attempt to give a unified presentation of Recursive Estimation and Identification. Approximately 3/4 of the course will be common to both subjects. The class will then be divided into two sections and the remainder of the course will concentrate on Continuous-time Estimation (for 6.433) and Identification Methods (for 6.435) respectively.

Topics:

Review of Discrete-time Stochastic Processes: stationarity, ergodicity

Review of Linear Systems Theory

Geometry of Linear Estimation; Estimation of Stochastic Processes

Models for Estimation and Identification with emphasis on State Space Models (Discrete-time)

Wiener and Kalman Filtering, Smoothing and Prediction

Parameter Estimation for Dynamical Systems: Prediction Error Formulation; Maximum

Likelihood Estimation

The above constitutes 3/4 of the course. The rest of the course proceeds as follows:

6.433 Recursive Estimation

Fast and Array Algorithms for Recursive Estimation

Continuous-time Wiener and Kalman Filtering

6.435 System Identification

Asymptotic Analysis of Predication Error Methods

Subspace Methods and Stochastic Realization Theory

Error Minimization vs. Complexity Tradeoff

Textbook:

Linear Estimation: T. Kailath, A.H. Sayed, B. Hassibi. Prentice Hall 2000.

Supplementary Notes: Notes by Sanjoy Mitter

Grades based on Homework and Term Paper

Time and Place:

The class will meet Monday and Wednesday from 2:30-4pm in 38-166.